You use two lock in amplifiers one driven by the sine , the other by the cosine. From the outputs you get amplitude and phase.
What ever it takes to cancel it. But we need to know the phase.
The PLL Bob was taking about will take care of the canceling amount but it would have to lock on to the harmonic.
Sure but you still have to tune it in. How to auto track?
I believe AD has a measurement IC (network analyzer) that might be able to do this.
The output from the sampler is IQ.
Last edited:
Seems like going backwards a bit but won't an out of balance dif stage generate hoards of 2ndH. The Tcell IC Bob used for the PLL might be ideal for this. In fact you can build a 4Q multiplier out them. $1.95 item.
A band pass filter might be ideal for rocking the phase. By tuning slightly above or below fundamental you can swing the phase +/- a few degrees. Might be able to do it with passives. The phase reverses at resonance at a rate proportional to Q.
Last edited:
do the cancellation with ADC/DSP/DAC?
much easier implement ultra narrow band PLL, Hilbert Transformers in DSP
even "only" -90 to -100 dB THD + Noise thru the processing won't hurt if correcting something already -100 dB
much easier implement ultra narrow band PLL, Hilbert Transformers in DSP
even "only" -90 to -100 dB THD + Noise thru the processing won't hurt if correcting something already -100 dB
The oscillator and the lockin amplifiers are synced to the same master clock.
Master clock has outputs at the fundamental and as many harmonics as you can cancel. Fundamental output is fed via gain control and filtered by just below oscillator resonator. That output resonator has variable gain/attenuator controlled by input feed and has as low as an output impedance as possible. The harmonic cancelators are attenuated and fed into the output. Then the lock in amplifiers check the harmonics and make the adjustments required. All "oscillators" are really D/As.
still thinking in big chunks ---- a bal output with 0 and -180 degree outputs.... and a pot to select anywhere between the two phase extreams.... used to keep in cancellation the harmonics in the output... PLL still needed.
THx-RNMarsh
Note I find several errors creeping into the measurements systems I have on hand from my personal free equipment rental supply...... 5 of them. each a different problem/solution. I did retune the generator for lowest thd on my system..... new Gen output is 1.93v rms.
The ranges of accuarcy in these low levels is the A-P 2722 (poor 2H recovery), and the ShibaSoku 725D (internal grounding/leakages) and then the best one - the unit made for Panasonic -- VP 7722A. The Panasonic seems to have addressed all the minor issues and annoyances of the others. Here is what I can get with it. Note no cancellation is being used external to the gen/analyzer --->
This is after a week of fussing around with grounding, isolation ETC and various analyzers residual funnies..... The Panasonic monitor output for 0dBV is -10dBv. I havent corrected for it in the photo.
THx-RNMarsh
Last edited:
When you start to get down this low in distortion, you have to look at the monitor output and resolve the remainder.... in my case thru an ADC/FFT (QA401).
THx-RNMarsh
THx-RNMarsh
I see a T branched BNC with one open end.
Does that affect signals/measurments?
Are there arrangements where we should never leave an open end?
Ed do you have a link to a lock-in amplifier that will lock to -100dB seconds in the presence of 0dB fundamental? You guys keep pulling out the sledge hammers, Bob's DM technique works like a charm.
Last edited:
Oh come on Scott were having fun.
Now will someone pass that really big sledge hammer over here.
Your answer "Ah, you need quadrature outputs to do that. "
My answer ....
Maybe I'm confused. Still waking up.
I used PLL in reference to the MC1496. I think he means 4Q multiplier.
Last edited:
Hi David,
Improved oscillators are out there cheaply these days. You can get a double oven type 10 MHz oscillator and set that up on a PCB. Feed it into your counter for an instant upgrade. The types that can be disciplined have a tuning voltage input, and there are DIY circuits out there that can take a 1 pps signal to control your master oscillator. That way you have the very low jitter of an x-tal oscillator, and the accuracy of a GPS locked source. A lot less expensive than the upgraded frequency reference option was at the time from HP (and even now from Keysight). If you can find a GPS timing receiver used, that and an antenna is all you need to completely solve your dilemma. Complete GPS timing solutions were in the $200 ~ $300 range, including timing antenna. There were some brand new systems for sale of a Lucent / Symetricom brand that had a receiver and back up frequency reference in case the primary reference went down. They used 5 MHz oscillators, but all you need is a frequency doubler to solve that issue. They also include a 1 pps output. You would need to provide a power supply (+24 VDC I think) to operate it. A gel cell would provide power over a short interruption. These use ovens, so the initial current draw is a few amperes, then drops to about 1/2 amp I think. I plan to run my heaters off a large gel cell, and the receivers (and oven electronics) from another supply to keep that stuff up all the time if I can. It will take a couple hours from a cold start for these systems to begin closing on a very accurate 10 MHz output. Sometimes as long as a week to get into the area where the frequency error is in the 10 exp-12 or so range. That is pretty darned close, enough for most of our uses. Only a time nut ( 🙂 ) would go for tighter accuracy (and they do). I'll also be running a stand alone HP ovenized oscillator for use as a sanity check. I can manually tune it every so often.
-Chris
Improved oscillators are out there cheaply these days. You can get a double oven type 10 MHz oscillator and set that up on a PCB. Feed it into your counter for an instant upgrade. The types that can be disciplined have a tuning voltage input, and there are DIY circuits out there that can take a 1 pps signal to control your master oscillator. That way you have the very low jitter of an x-tal oscillator, and the accuracy of a GPS locked source. A lot less expensive than the upgraded frequency reference option was at the time from HP (and even now from Keysight). If you can find a GPS timing receiver used, that and an antenna is all you need to completely solve your dilemma. Complete GPS timing solutions were in the $200 ~ $300 range, including timing antenna. There were some brand new systems for sale of a Lucent / Symetricom brand that had a receiver and back up frequency reference in case the primary reference went down. They used 5 MHz oscillators, but all you need is a frequency doubler to solve that issue. They also include a 1 pps output. You would need to provide a power supply (+24 VDC I think) to operate it. A gel cell would provide power over a short interruption. These use ovens, so the initial current draw is a few amperes, then drops to about 1/2 amp I think. I plan to run my heaters off a large gel cell, and the receivers (and oven electronics) from another supply to keep that stuff up all the time if I can. It will take a couple hours from a cold start for these systems to begin closing on a very accurate 10 MHz output. Sometimes as long as a week to get into the area where the frequency error is in the 10 exp-12 or so range. That is pretty darned close, enough for most of our uses. Only a time nut ( 🙂 ) would go for tighter accuracy (and they do). I'll also be running a stand alone HP ovenized oscillator for use as a sanity check. I can manually tune it every so often.
-Chris
Since I work for a cable services company. The cable and connectors are free.
You mentioned running the oven off cells. Is that the complete unit or does the supply to just the heating system effect things?
I think it's just a stable temperature that is required. But you are expert.
Last edited:
What amount of frequency accuracy is required for a low distortion RC gen.
I thought the idea is to satisfy FFT requirements to minimize the know issues.
One can easily over design such a system.
I thought the idea is to satisfy FFT requirements to minimize the know issues.
One can easily over design such a system.
No problems, I can't help but point out a sine and its inverse can not be combined in any proportion to make a cosine, let alone the double inverse (-180) which is 0.